Water Treatment Capacity of Forward-Osmosis Systems Utilizing

May 15, 2015 - ... Engineering and ‡Department of Engineering and Public Policy, .... Tao Yan , Yuanyao Ye , Hongmin Ma , Yong Zhang , Wenshan Guo ,...
0 downloads 4 Views 1MB Size
Article pubs.acs.org/IECR

Water Treatment Capacity of Forward-Osmosis Systems Utilizing Power-Plant Waste Heat Xingshi Zhou,† Daniel B. Gingerich,‡ and Meagan S. Mauter*,†,‡ †

Department of Chemical Engineering and ‡Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, United States S Supporting Information *

ABSTRACT: Forward osmosis (FO) has the potential to improve the energy efficiency of membrane-based water treatment by leveraging waste heat from steam electric power generation as the primary driving force for separation. In this study, we develop a comprehensive FO process model, consisting of membrane separation, heat recovery, and draw-solute regeneration (DSR) models. We quantitatively characterize three alternative processes for DSR: distillation, steam stripping, and air stripping. We then construct a mathematical model of the distillation process for DSR that incorporates hydrodynamics, mass- and heattransport resistances, and reaction kinetics, and we integrate this into a model for the full FO process. Finally, we utilize this FO process model to derive a first-order approximation of the water production capacity given the rejected-heat quantity and quality available at U.S. electric power facilities. We find that the upper bound of FO water treatment capacity using low-grade heat sources at electric power facilities exceeds process water treatment demand for boiler water makeup and flue-gas-desulfurization wastewater systems.



INTRODUCTION Proposed effluent limitation guidelines at steam electric powergeneration facilities will significantly increase the demand for on-site water treatment.1 One opportunity to minimize the auxiliary power consumption associated with this treatment capacity is to utilize waste heat available on-site for membranebased water treatment. One potential technology is forward osmosis (FO), where the draw solution is a thermolytic salt (e.g., NH4HCO3) (Figure 1).2−4 In this two-step process, feedwater is drawn across a semipermeable membrane by a difference in osmotic pressure between the feed solution and the draw solution. The dilute draw solute is then regenerated by thermal decomposition of the thermolytic salt into its constituent gases (i.e., NH3 and CO2).5 If waste heat is available, this separation process offers significant electricity savings over reverse osmosis.6,7 Steam electric power-generation facilities are the largest source of waste heat in the United States,8 but the feasibility of utilizing this waste heat to drive FO separation processes has yet to be systematically assessed in the peer-reviewed literature.5,6,9 Past modeling efforts to evaluate the feasibility of waste-heat-driven FO assumed that heat is available at desired quantities and temperatures,7 whereas experimental demonstrations of FO processes in the peer-reviewed literature utilized electricity or fuel to generate heat.9−11 This significant gap in the literature exists largely because robust estimates of the quantity, quality, and availability of power-plant waste heat are sparse.12,13 Our recent work provides estimates of the quantity, quality, and spatial-temporal availability of waste heat for the U.S. power sector over the next 30 years.14 Demonstrating the feasibility of power-plant waste-heat-driven FO requires the integration of these waste-heat estimates with heat capture, transport, draw-solute regeneration (DSR), and membrane separation models. © XXXX American Chemical Society

Accurate modeling of the DSR system also depends on a robust understanding of draw-solute chemistry. Much of the past modeling work failed to report the specific method for evaluating the separation performance of the CO2−NH3−H2O ternary system.7,9,15 Other models underestimated heat consumption per unit of product water or column height by assuming that the system reaches equilibrium over the column height. For example, Kim et al.’s equilibrium-based simulations underestimated the heating energy by 59% compared to experimental data from their pilot recovery process.16 Recent efforts to model the ammonia-based CO2 capture process provided a potential route for including the influence of hydrodynamics, mass- and heat-transport resistances, and reaction kinetics into the DSR model.17−20 The application of these rate-based models to the FO draw-solute-regeneration system ensures the validity of the DSR process simulations, but these rate-based models are also computationally intensive and, therefore, unfavorable for optimization case studies. To overcome this issue, we develop both a rate-based model and a reduced model that simplifies the rate-based model with negligible deviation from predicted values. Finally, optimizing the water production capacity of wasteheat-driven FO systems has been hampered by reliance on the ASPEN modeling environment. ASPEN, and associated blackbox optimization techniques, pose numerical issues (e.g., system convergence, initialization, nonconvexity), are timeconsuming, and do not efficiently perform multiobjective optimization.7,9 When optimization of DSR systems was performed by past researchers, the objective was to minimize Received: February 4, 2015 Revised: May 10, 2015 Accepted: May 15, 2015

A

DOI: 10.1021/acs.iecr.5b00460 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Figure 1. Generalized process flow diagram of the forward-osmosis process.

the cost of the system rather than to maximize the amount of water constrained by a certain heat input.7 The distinction is particularly important for systems utilizing waste heat as the energy source. Integrated mathematical models of FO unit processes will ultimately allow for efficient optimization and a thorough characterization of the trade-offs between operating conditions and cost. This work evaluates the feasibility of using waste heat rejected from steam electric power-generation processes to treat water of moderate salinity [∼35000 ppm total dissolved solids (TDS)], such as flue-gas-desulfurization wastewater, onsite. In 2012, coal and natural gas generators discharged 900 million GJ of heat to the environment in their exhaust streams.14 This heat has temperatures ranging from an average of 90 to 543 °C, depending on the fuel cycle [i.e., coal, natural gas combined cycle (NGCC), natural gas steam turbine (NGST), and natural gas gas turbine (NGGT)]. Using simulation models, we identify the most energy-efficient DSR design given the heat quality available at the median U.S. power plant for each fuel cycle. We then develop a mathematical model of the complete FO process, including the membrane separation step, the heat recovery step,14 and the DSR process, to evaluate trade-offs between various operating conditions. Finally, we apply solutions from the mathematical model to determine the upper bound of water volume that could feasibly be treated by FO using waste heat from the flue-gas streams of power plants. We limit our water production capacity analysis to waste heat available on-site, as this waste heat has commonly been cited as an ideal source of energy for FO processes.6,21



METHODOLOGY

CO2−NH3−H2O Separation Models in DSR Systems. In this section, we model regeneration of the thermolytic CO2− NH3 draw-solute system. In aqueous solution, complex reversible reactions occur [see Table S1 of section S1 of the Supporting Information (SI) for the list of these reactions], and species including bicarbonate, carbonate, and carbamate are formed.19,20,22 Although Table S1 (SI) lists a precipitation/ dissolution reaction (R6), we neglect this reaction in the present work because the maximum concentration of the draw solution, 8 M (CO2-based), is far below the solubility limit of 13 M. Previous work in modeling this ternary system for CO2 capture suggested that hydrodynamics, thermodynamics, ratebased mass and heat transfer, and reaction kinetics are critical to accurately capturing the energy intensity of the regeneration process.20 We start modeling the DSR system by developing a rate-based model that considers hydrodynamics, thermodynamics, rate-based mass and heat transfer, and reaction kinetics. Next, we simplify the rate-based model to develop an equilibrium-stage model that considers thermodynamics and reaction kinetics but neglects mass-transport resistance and heat-transport resistance. Given a distillation column of infinite residence time, the rate-based model and equilibrium model produce very similar estimates of normalized heat duty. Finally, we develop a basic chemistry model that assumes instantaneous chemical equilibrium and neglects hydrodynamics and interfacial mass- or heat-transport resistance in the column. B

DOI: 10.1021/acs.iecr.5b00460 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

pressure of volatile components in the gas phase, as well as increasing the temperature. According to Henry’s law, fractions of the volatile components in the liquid phase decrease because of vapor−liquid equilibrium limitations, leading to the separation of CO2 and NH3 from the clean water. We build models for the three alternative DSR processes in ASPEN Plus (Aspen Technology, Inc.) using the electrolyte nonrandom two-liquid Redlich−Kwong (NRTL-RK) thermodynamic method. We specify the flow rate and concentration of the dilute draw solution, the number of stages, the boilup ratio, and the packing. ASPEN Plus simulates the required heat duty, as well as the column diameter, concentration and flow rate of the product water, and concentration and flow rate of the concentrated draw solution. The rate-based model takes hydrodynamic, thermodynamic, rate-based heat/mass transfer, and chemical reaction into account, as shown in Table S2 of section S2 of the SI, on the equipment specifications for the ASPEN model. To perform rate-based simulations, the columns are specified to contain Amistco (Goodloe equivalent) structured packing4 with a void fraction of 0.945, a specific area of 580 ft2/ft3, a static holdup of 5%, and a pressure drop of 0.096 mmHg/ft. Packing height is approximately 8.3 ft (equivalent to 25 theoretical stages). We vary the column pressure depending on the specific case (i.e., ambient or subatmospheric). Thermal and electrical energy requirements are calculated based on a product-water quality specified to contain less than 1 ppm of ammonia (including related species such as ammonium and carbamate). The equilibrium-stage model was constructed by switching from “rate-based” mode to “equilibrium” mode in ASPEN Plus and disabling rate-based modeling in the packing rating section. The chemistry model was constructed by further deleting the reaction section in the column specifications, thereby leveraging the chemistry model without the consideration of rate-based heat/mass transfer and reaction kinetics. We quantitatively evaluate each DSR method by determining its normalized heat duty in ASPEN Plus. Considering the relatively low energy consumption (as detailed in the first two subsections of the Results and Discussion section), potential issues with the steam contaminating the product water in steam stripping processes, and the option of accessing lower-quality heat by operating the distillation column at subatmospheric pressures, we selected distillation as the optimal DSR system for mathematical model formulation. Mathematical Model of Distillation for a DSR System. The majority of energy consumption of the FO desalination process lies in the DSR system, and further evaluation and optimization of this process in the context of other FO system components is greatly aided by a rigorous mathematical model of the DSR process. We formulate a rate-based model for drawsolute recovery based on a comprehensive MERSHQ model20 including mass balances (M), energy balances (E), rate equations (R), summation equations (S), hydrodynamic equations (H), and equilibrium equations (Q) as detailed below. This model assumes that stages are well-mixed and have similar properties (e.g., temperature) throughout the stage as calculated within the model. We model the distillation column to have metal gauze structure packing in the “X” configuration23 to minimize the pressure drop in the column. Mass Balances. Material balances in the vapor (MVi,j) and liquid (MLi,j) phases for component i at stage j depend on the flow rates of the vapor (Vj), liquid (Lj), and feed (Fj); the mole fractions in the vapor (yi,j) and liquid (xi,j); the molar flux (Ni,j);

The resulting models and their major differences are listed in Table 1. Table 1. Model Comparison model name

thermodynamics

reaction kinetics

hydrodynamics

rate-based mass/heat transfer

chemistry model equilibriumstage model rate-based model

yes

no

no

no

yes

yes

yes

no

yes

yes

yes

yes

For all models, we report computational speed and compare heat duty estimates in Table S7 of section S10 of the SI. Although we expect the greatest accuracy from the rate-based model, incorporation of hydrodynamics and rate-based mass and heat transfer increases the complexity of the models and significantly decreases the computational speed of the calculations. Therefore, we use the solution from the ratebased model as a reference for evaluating the validity of the basic chemistry and equilibrium-stage models. Evaluating Separation Processes for DSR. Although selection of an appropriate DSR process depends on the properties of the draw solute, the efficiency of upstream membrane processes, size or process constraints imposed by site conditions, cost, and many other factors, in this initial study, we consider only the limitations imposed by heat quality and quantity available at electric power-generation facilities. The reversible thermal decomposition of the species in the dilute draw solution, including ammonium bicarbonate, ammonium carbonate, and ammonium carbamate, into CO2 and NH3 can be achieved by either the addition of heat or the reduction of partial pressure. As a result, both conventional distillation and stripping techniques are potentially viable means of regenerating the draw solution. We consider three alternative processes, namely, distillation, steam stripping, and air stripping, to separate the dilute draw solution into a concentrated NH3/CO2 stream and a low-salinity product water that meets drinking water standards for ammonia of 1 mg/L. An overview of the DSR process is provided in Figure 1. We model each of the DSR processes using a packed column. The DSR process commences with preheating of the dilute draw solution, utilizing the thermal energy of hot streams from the top or bottom of a packed column. The dilute draw solution at or above its bubble point, depending on the specific operating pressure, then enters the packed column, where one of the three alternative processes is performed. In the distillation-based DSR process, a reboiler at the bottom of a conventional distillation column provides the thermal input for distillation. The indirect heating of distillation is likely to be less efficient than stripping methods because of the thermal inefficiency introduced by the heat exchanger, but this process is better suited for low-temperature DSR, where subatmospheric pressures are required to completely recover the draw solution. In the steam stripping process, the latent heat of steam directly heats the sump streams in the column. This process is hindered by steam purity, as condensation can result in impurities in the product water and affect its downstream application. The final DSR method considered, hot air stripping, drives the separation by reducing the partial C

DOI: 10.1021/acs.iecr.5b00460 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research Hj = Pj − Pj − 1 − ΔPj − 1 = 0

the effective interfacial area (ae); the volume of column per stage (VS); and reactions that occur that depend on the stoichiometric coefficients (ν), the reaction rates (r), and the reaction extent (X) for each of the controlled reactions (the number of which is given by NRC) and instantaneous reactions (the number of which is given by NRE) MiV, j = Vjyi , j − Vj + 1yi , j + 1 − FjVyiF, j − NiV, j ae(VS)j = 0

Equilibrium Equations. The vapor−liquid equilibrium at the interface (QIi,j) is a function of the vapor−liquid equilibrium constant (Ki,j) and the mass fractions in the liquid and vapor phases at the interfaces Q iI, j = K i , jxiI, j − yiI, j = 0

(1)

NRC

NRE

k=1

∑ (νi′,kχj ,k )] = 0 k=1

(2)

Details on the calculation of mass- and heat-transfer behaviors can be found in section S3 of the SI. Information on the activity correction using the electrolyte NRTL model can be found in section S4 of the SI. The total material balance equations at stage j are given by M T,V j = Vj − Vj + 1 − FjV + NT,V jae(VS)j = 0

(3)

M T,L j = Lj − Lj − 1 − F jL − NTL, jae(VS)j − (VLH)j C

NRC

NRE

∑ [ ∑ (νi ,krj , k) + ∑ (νi′,kχj ,k )] = 0 i=1

k=1

k=1

(4)

Mi , j = Ljxi , j + Vjyi , j − Lj − 1xi , j − 1 − Vj + 1yi , j + 1 − FjVyiF, j

Energy Balances. Energy balance equations for the vapor (EVj ) and liquid (ELj ) phases and the interface (EIj ) at stage j depend on the vapor, liquid, and feed flow rates; the molar enthalpy (H); the heat duty added at the each stage (Q); and the interfacial energy (ϵ) EjV = VjHjV − Vj + 1HjV+ 1 − FjVHjVF + Q jV + ϵVj = 0

(5)

EjL = LjHjL − Lj − 1HjL− 1 − F jLHjLF + Q jL − ϵLj = 0

(6)

NRC



R iL, j = Ni , j − NiL, j = 0

(9)

i=1

∑ xiI,j − 1 = 0 i=1

(15)

The summation equations for both the liquid and vapor phases, which are functions of the molar fractions in the vapor (yi,jI ) and liquid (xIi,j) phases, are expressed as C

SjL =

C

∑ xi ,j − 1 = 0, i=1

SjV =

∑ yi ,j − 1 = 0 i=1

(16)

Vapor−liquid equilibrium still holds, but the equilibriumstage model replaces the interfacial composition with the compositions in the bulk of the liquid (xi,j) and vapor (yi,j) phases Q i , j = K i , jxi , j − yi , j = 0

(17)

Supplemental Equations for Chemical Reactions. The expressions for the equilibrium constants (Kj,k) of the instantaneous reactions are function of the temperature in the stage (Tj) and are considered in the rate-based, equilibrium, and chemistry models

(10)

C

SjIL =

k=1

− FjVHjVF + Q j = 0

C

∑ yiI,j − 1 = 0

∑ (νi′,kχj ,k )] = 0

Ej = LjHjL + VjHjV − Lj − 1HjL− 1 − Vj + 1HjV+ 1 − F jLHjLF

Summation Equations. The summation equations for the IL vapor phase (SIV j ) and liquid phase (Sj ) are functions of the I molar fraction in the vapor (yi,j) and liquid (xIi,j) phases SjIV =

− (VLH)j [ ∑ (νi , krj , k) +

The energy balances at stage j, which depend on the vapor, liquid, and feed flow rates; the molar enthalpy (H); and the heat duty added at the each stage (Q), are expressed as

Rate Equations. The reaction rate for each component i on each stage j in the vapor (RVi,j) and liquid (RLi,j) is a function of the molar flux for each component i on each stage j in the vapor (NVi,j) and liquid (Ni,jL ) (8)

NRE

(14)

(7)

R iV, j = Ni , j − NiV, j = 0

F jLxiF, j

k=1

The energy balance equation at the interface is EjI = ϵVj − ϵLj = 0

(13)

Similarly to the approach that we took in developing the ASPEN models, we develop both the chemistry model and the equilibrium-stage model as simplified versions of this original rate-based model. In the chemistry model, we assume instantaneous chemical equilibrium for the reactions by neglecting all kinetics terms and replacing the interfacial composition with the composition in the bulk phase. In the equilibrium-stage model, we do account for kinetics, yielding the sets of equations listed below. The material balances in the vapor (Mi,jV ) and liquid (Mi,jL ) phases for component i at stage j depend on the flow rates of the vapor (Vj), liquid (Lj), and feed (Fj); the mole fractions in the vapor (yi,j) and liquid (xi,j); the volume of column per stage (V S ); and reactions that occur that depend on the stoichiometric coefficients (ν), the reaction rates (r), and the reaction extent (X) for each of the controlled reactions (NRC) and instantaneous reactions (NRE)

MiL, j = Ljxi , j − Lj − 1xi , j − 1 − F jLxiF, j − NiL, jae(VS)j − (VLH)j [ ∑ (νi , krj , k) +

(12)

(11)

Hydraulic Equations. The hydraulic equation at each stage (Hj) is a pressure balance equation, so that, at any stage j, the pressure (Pj) minus the pressure drop in each stage (ΔPj−1) is equal to the pressure on the next stage (Pj−1)

ln K jeq, k =

Ak + Bk ln Tj + CkTj + Dk = Tj

(k = 1, NRE) D

C

∑ νi′,k ln ai ,j i=1

(18) DOI: 10.1021/acs.iecr.5b00460 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

where t, τ, and ε express the thickness, tortuosity, and porosity, respectively, of the support layer. The mass-transfer coefficient is correlated with the Sherwood number and is calculated as

The reaction rates (rj,k) of the rate-controlled reactions are functions of the temperature, the energy of activation (Ek), the ideal gas constant (R), the mass fraction in the liquid, the activity coefficient (γi,j), and the liquid-phase activity (ai,k), but are considered in only the rate-based model rj , k

⎛ E ⎞ C a, k ⎟⎟ ∏ (xi , jγi , j)ai ,k = kTj n exp⎜⎜ − RT ̅ ⎝ j ⎠ i=1

κ=

(k = 1, NRC)

The values for equilibrium constants and reaction rates are listed in Tables S3 and S4, respectively, of section S5 of the SI. The CO2−NH3 system has mixed polar and nonpolar components in its vapor phase. To account for this and the nonideality of our system, we estimate the fugacities, activities, and densities in the system using the methods listed in Table S5 (section S6, SI). Table S5 (SI) details the parameter estimation methods used in the current work, whereas Table 2 details the input parameters for the simulation study of the DSR. We allow the draw solution entering the column to range in concentration from 0.5 to 1.5 M (CO2-based).

J1 =

J2 =

Table 2. Specifications of the DSR System in this Case Study value in the case study

dilute DS column feed (m3/h) C/N ratio of draw solution column pressure (atm) feed concentration (M) target NH3 content in product water

18 0.714 1 1 680.9

a

DDS flow rate =18 m3/h.

Figure 5. Parametric analyses of the heat duty of the FO process under various operating parameters. (A) Parametric analysis of the effect of the distillation column operating pressure on the energy consumption of a distillation column with 25 stages with a 1 M column feed, target NH3 concentration of 1 ppm, and a C/N ratio of 0.714. (B) Parametric analysis of the target water purity for columns of different sizes on energy consumption for a 1 M column feed, at a pressure of 1 atm and a C/N ratio of 0.714. (C) Parametric analysis of the effect of the dilute-draw-solution feed concentration on the energy consumption of a distillation column with 25 stages with a pressure of 1 atm, a target NH3 concentration of 1 ppm, and a C/N ratio of 0.714. (D) Parametric analysis of the effect of the C/N ratio of the draw solution for the membrane process on the energy consumption of a distillation column with 25 stages operating at a pressure of 1 atm and with a target NH3 concentration of 1 ppm. H

DOI: 10.1021/acs.iecr.5b00460 Ind. Eng. Chem. Res. XXXX, XXX, XXX−XXX

Article

Industrial & Engineering Chemistry Research

Resulting water production rates for U.S. power plants by fuel cycle are reported in Figure 6. Figure 6A depicts the

heat quality (Figure 5A). However, a lower pressure also leads to a higher heat duty per unit volume of clean water produced, as discussed in the Comparison of DSR Processes section. Column Height. The column size also influences the energy consumption of the process. We analyze the sensitivity of the heat duty to the target ammonia concentration in the product water for columns of three different heights (Figure 5B). We find that, given constraints of